Neuronpedia logo - a computer chip with a rounded viewfinder border around it

    Neuronpedia

    APICircuit TracerNEWSteerSAE EvalsExportsSlackBlogPrivacy & TermsContact
    © Neuronpedia 2025
    Privacy & TermsBlogGitHubSlackTwitterContact
    1. Home
    2. Google DeepMind · Exploring Gemma 2 with Gemma Scope
    3. Gemma-2-2B
    4. Attention Out - 16k
    5. 10-GEMMASCOPE-ATT-16K
    6. 5829
    Prev
    Next
    INDEX
    Explanations

    attends to tokens that are members or properties of objects from tokens that are method calls or identifiers associated with those objects

    oai_attention-head · gpt-4o-miniTriggered by @bot
    New Auto-Interp
    Top Features by Cosine Similarity
    Comparing With GEMMA-2-2B @ 10-gemmascope-att-16k
    Configuration
    google/gemma-scope-2b-pt-att/layer_10/width_16k/average_l0_70
    Prompts (Dashboard)
    36,864 prompts, 128 tokens each
    Dataset (Dashboard)
    monology/pile-uncopyrighted
    Features
    16,384
    Data Type
    float32
    Hook Name
    blocks.10.attn.hook_z
    Hook Layer
    10
    Architecture
    jumprelu
    Context Size
    1,024
    Dataset
    monology/pile-uncopyrighted
    Activation Function
    relu
    Embeds
    IFrame
    Link
    Not in Any Lists

    No Comments

    Head Attr Weights
    0:0.26
    1:0.10
    2:0.17
    3:0.05
    4:0.07
    5:0.15
    6:0.06
    7:0.11
    Negative Logits
    stoke
    -0.31
    FormTagHelper
    -0.30
    gih
    -0.28
    var
    -0.28
    dist
    -0.28
     is
    -0.28
    Dist
    -0.27
    IM
    -0.27
    is
    -0.26
     zak
    -0.26
    POSITIVE LOGITS
    osoever
    0.42
     Audiodateien
    0.41
    perrt
    0.39
    berfläche
    0.39
     AssemblyTitle
    0.38
    tonode
    0.38
     Sandoval
    0.37
    AutoScale
    0.37
     resourceCulture
    0.36
    سكرية
    0.36
    Activations Density 0.046%

    No Known Activations